MetaTrader is a moat, not a product — and AI is dissolving it.

Ask a broker CEO why they run MT4 or MT5 and you will hear a version of the same answer. It is what traders know. It is what the affiliate ecosystem was built on. It is what every EA developer targets. It is where the flow lives. None of those answers is about the product. All of them are about the switching cost. MetaQuotes is not a technology company competing on product. It is a switching-cost company competing on inertia.

For twenty years that was a great business. Inertia is stable. Inertia compounds. Every new broker that launched on MT4 deepened the moat for every other broker who was already there. The problem for MetaQuotes is that inertia has one enemy — and that enemy just showed up.

What AI actually collapses

Switching costs in a retail trading platform come from three places. Muscle memory — the trader knows where every button is. Ecosystem — indicators, EAs, plugins, affiliate tools. And integration — the broker’s middle office is wired into the platform.

AI collapses two of the three in a specific way. Muscle memory becomes irrelevant when the platform’s primary interface is natural language over intent. A trader who used to hunt through nested menus now says find me EUR pairs with tight spreads in the London session and gets an answer. The muscle memory advantage evaporates because there is no muscle involved.

Ecosystem collapses the moment a brokerage’s bespoke platform can translate MT-family EAs into its own execution environment on the fly. That translation is now a solved engineering problem. It is not zero cost, but it is not the two-year rewrite it used to be. Integration is the last remaining moat, and it is the one middle-office engineering can dismantle in a quarter.

What replaces it

The bespoke brokerage platform of 2026 is not a MetaTrader clone. It is a substrate. Execution engine at the base — usually an existing one, sometimes bought, sometimes reused. On top of that, a policy layer, a retrieval layer, and an evals layer identical in shape to the three-layer architecture we ship for compliance. On top of that, an intent-first interface that traders drive by describing what they want rather than by clicking through menus.

The competitive point is not that this platform is better than MT5. It is that the brokerage owns it. Every feature is theirs. Every workflow is theirs. Every UX experiment is theirs. Every trader interaction becomes proprietary data that improves the platform for the next trader. None of that is possible on a licensed shared platform.

The build cost of a serious bespoke platform is real — usually £750,000 to £2 million for the first release — but it replaces a MetaQuotes bill that runs into six figures per month at scale and is spent, not invested. Two years in, the ROI comparison stops being close.

The five moats brokerages should be building instead

The bespoke platform is not the point. It is the vehicle. The point is the moats that can only be built on top of a substrate you own.

• A retention copilot that watches trader behaviour and intervenes before the churn window opens.
• A KYC and onboarding stack that closes accounts in minutes rather than days.
• A dealer-desk assistant that summarises risk positions and flags anomalies in real time.
• A regulatory reporting layer that produces evidence artefacts automatically for CySEC, FCA, ASIC, whichever the broker sits under.
• A voice-of-trader system that turns support conversations into product-priority signal — the single most underused data asset in every brokerage we have seen.

The transition path

No serious brokerage should rip and replace on day one. The pattern that works is dual-run. Keep the existing MT-family deployment. Build the bespoke substrate alongside it. Migrate one trader cohort at a time, starting with the segment that generates the most support cost and the least revenue — traders who cost you money to keep on the old platform. Measure retention, CSAT and lifetime value against the control group.

In every dual-run migration we have observed, the new-platform cohort outperforms on all three within a quarter. That is the point at which the CFO signs off on the full migration. Not before. Do not try to sell the CFO on the concept. Sell them on the cohort data.

The affiliate problem

The affiliate objection is real. A big part of the retail broker economy depends on affiliates who built landing pages, funnels and communities around MT-family platforms. Migrating away creates real friction there.

The response is not to fight the objection but to reprice it. A bespoke platform can offer affiliates a per-cohort custom experience — branded builds, custom indicators, custom dashboards, custom promotions — that no shared platform can. The affiliates that matter will move for that. The ones that will not are, in most cases, not the ones producing the LTV that justifies keeping the old stack.

What the next 24 months look like

By late 2027, the top decile of retail brokerages will be running on bespoke substrates. The middle 60% will still be on MT-family platforms, watching their per-trader economics quietly worsen because they cannot build any of the moats above. The bottom decile will consolidate into the middle 60% as capital dries up.

The MT-family install base will not disappear. It will become the plumbing of a shrinking middle. And MetaQuotes, which for twenty years priced its product like a moat because it was one, will discover that moats stop earning premium the moment a viable competitor arrives.

For a broker executive reading this, the honest question is not whether to move but when. Every quarter on the shared platform is a quarter your competitor spends building a substrate you cannot see yet.